Intel Corporation

Donald D Parker

Gary L Brown

Robert J Greiner

Stephen S Pawlowski

David Lawrence Hill

Gurbir Singh

Darrell D Boggs

Adrian L Carbine

Michael M Hancock

Alan Beecher Kyker

Rajesh Kumar

Bradley D Hoyt

Gail M Rupnick

A bus agent that may be used in an enhanced highly pipelined bus architecture. In one embodiment, the bus agent includes a control interface to drive a control signal at a clock frequency, an address bus interface to drive address elements at twice the clock frequency, and a data bus interface to drive data elements at four times the clock frequency. The address bus interface drives a substantially centered address strobe transition for each address element, and the data bus interface drives a substantially centered data strobe transition for each data element.

Enhanced highly pipelined bus architecture

A bus agent that may be used in an enhanced highly pipelined bus architecture. In one embodiment, the bus agent includes a set of snoop status interfaces, an address strobe signal interface, and a bus clock interface for a bus clock signal. The bus agent of this embodiment also includes bus controller logic capable of sensing or asserting one or more of a set of snoop status signals for transaction N on the snoop status interfaces during a snoop phase to start in a bus cycle upon the later of three or more bus clock cycles of the bus clock signal after a beginning of a bus cycle of an the assertion of an address strobe signal for transaction N or two or more bus clock cycles of the bus clock signal after a beginning of a bus cycle in which a most recent snoop phase begins.

Snoop phase in a highly pipelined bus architecture

A bidirectional multidrop processor bus is connected to a plurality of bus agents. Bus throughput can be increased by operating the bus in a multi pumped signaling mode in which multiple information elements are driven onto a bus by a driving agent at a rate that is a multiple of the frequency of the bus clock. The driving agent also activates a strobe to identify sampling points for the information elements. Information elements for a request can be driven, for example, using a double pumped signaling mode in which two information elements are driven during one bus clock cycle. Data elements for a data line transfer can be driven, for example, using a quad pumped signaling mode in which four data elements are driven during one bus clock cycle. Multiple strobe signals can be temporarily activated in an offset or staggered arrangement to reduce the frequency of the strobe signals. Sampling symmetry can be improved by using only one type of edge (e.g., either the rising edges or the falling edges) of the strobe signals to identify the sampling points.

Quad pumped bus architecture and protocol

A bus agent that may be used in an enhanced highly pipelined bus architecture. In one embodiment, the bus agent includes a target ready interface, a set of response interfaces for a set of response signals, and a data bus busy interface, and a bus clock interface for a bus clock signal. The bus agent of this embodiment also includes bus controller logic to track a plurality of transactions comprising a transaction N- and a transaction N, the bus controller being capable of asserting the target ready signal for transaction N if the bus agent is asserting the data busy signal for the transaction N- and deasserts the data busy signal.

Response and data phases in a highly pipelined bus architecture

A method and apparatus for decoding a macroinstruction, the macroinstruction including an operational code (opcode) and a specification of an operand, is described. The method includes two primary steps, which are performed either serially or in parallel. When performed serially, the steps may be performed in any order. The first primary step involves the generation of a first micro-instruction, specifying a first micro-operation, the first micro-instruction being derived from the specification of the operand of the macroinstruction. The second primary step involves the generation of a second micro-instruction, specifying a second micro-operation, the second micro-instruction being derived from the opcode of macroinstruction. The specification of the operand may specify the operand to be either a memory operand or a register operand in a manner that necessitates data processing or manipulation prior to a memory access or to execution of the second micro-instruction. More specifically, the specification of the operand in the macroinstruction may require alignment of operands retrieved from registers, prior to execution of the second instruction. In this case, the first micro-instruction may require a shift operation after retrieval of the operands.

Method and apparatus for generating a microinstruction responsive to the

A circuit and method for supplying and aligning a block of multiple variable length macro instructions to an instruction buffer. Only one cycle is required to align and rotate the block of instruction code. A last byte vector of the instruction code in the instruction buffer is scanned from the last byte in a direction from back to front, thereby saving time. Rotating begins immediately so that a next block of instruction code is available in a next cycle. The block is stored in the instruction buffer after all macroinstructions therein have been steered to the decoder.

Method and apparatus for aligning an instruction boundary in variable

A micro-operation queue for holding a plurality of micro-operations supplied simultaneously by a decoder. A plurality of packing multiplexers are coupled to receive the plurality of micro-operations, and valid bits associated therewith, and to provide packed micro-operation data output in which the valid micro-operations are positioned in adjacent outputs, thereby removing all empty slots. A FIFO queue receives the packed data, in responsive to valid micro-operations, stores the valid micro-operations starting with the next available empty queue location. An embodiment described in which the FIFO queue includes a circular queue with a plurality of entries. In one embodiment, alignment multiplexers for the circular queue are combined with the packing multiplexers, to provide a single-level plurality of packing and aligning multiplexers that has a control system that, responsive to the valid bits of the packed data and the next available pointer of the circular queue, packs, aligns, and stores the micro-operations into the circular queue from where they can be issued.

Packing valid micro operations received from a parallel decoder into

A split queue system for a decoder that supplies one or more micro-operations and data associated with the micro-operations. A main queue is coupled to receive one or more micro-operations from the decoder, and supply it to a next processing stage to provide a process micro-operation. A shadow queue is coupled to receive data associated with the micro-operation, in the same cycle that the micro-operation is supplied to the main queue. A control circuit is coupled to the main queue for issuing micro-operation from the main queue into the next processing stage in a first cycle, and in a second cycle issuing, the micro-operation therefrom. Also in the second cycle, the control circuit issues the data associated with the micro-operation from the shadow queue, so that the data is synchronized with its associated processed micro-operation.

Decoder having a split queue system for processing intstructions in a

A decoder for decoding multiple instructions in parallel, including a full decoder that can decode an instruction into multiple micro-operations, and a partial decoder that can decode a subset of the full instruction set.

Decoder for decoding multiple instructions in parallel

A circuit and method for simultaneously steering multiple aligned macroinstructions from an instruction buffer to a decoder that receives and decodes multiple macroinstructions in parallel. A first macroinstruction is supplied to a first decoder by steering a first predetermined number of bytes following the first buffer byte. A second macroinstruction is supplied by scanning a first opcode byte vector to locate a first opcode byte, and then steering a second predetermined number of bytes beginning at said first opcode to a second decoder. Operations to locate the first byte of each of the macroinstructions and to steer them to the decoders are accomplished in one cycle. If said macroinstruction cannot be decoded by said second decoder, then it is resteered to the first decoder. Steering and resteering operations continue until all complete macroinstructions within the instruction buffer have been accepted by the decoders.

Method for parallel steering of fixed length fields containing a

Decoding circuitry and a method supplying an immediate field that is issued from a decoder. A macroinstruction is supplied to the decoding circuit, which generates a first micro-operation that includes a first aliasing field and a first immediate field. The first aliasing field indicates the source of the micro-operation that will eventually be issued from the decoder. If the source is the first immediate field, then the alias field is further examined to determine the interpretation to be placed upon the data. The data may be interpreted literally, or as an address into a constant ROM, thereby providing an ability to output wide, 32-bit immediate data from a narrower, 9-bit input addresses. Additional sources for immediate data include macro-alias registers, macro-branch information, and micro-branch information.

Decoding circuit and method providing immediate data for a

A state recovery and restart method that simplifies assist handling. The recovery and restart method also handles micro-branch mispredictions. An assist sequence is executed in microcode to assist an error-causing macroinstruction. If data is required from an error-causing macroinstruction, it is fetched, decoded, and macro-alias registers are restored with macro-alias data. To recover the state of the micro-alias registers, micro-alias data from a micro-operation of the flow may be loaded into the micro-alias register. Subsequently, control returns to the Micro-operation Sequence (MS) unit to issue further error correction Control micro-operations (Cuops). In order to simplify restart, the Cuops originating from the error-causing macroinstruction supplied by the translate programmable logic arrays (XLAT PLAs) are loaded into the Cuop registers, with their valid bits unasserted. If microcode requests a restart beginning at one of the Cuops stored in the Cuop register, then the bits for that Cuop and subsequent Cuops are marked valid. Thus, the instruction can be restarted anywhere within the microcode sequence.

Growing community of inventors

Portland, OR, United States of America

Donald D Parker